# October 2026 data and exhibit notes

## Reconciled source scope

- The source package contains 50 timestamped event exports and a consolidated Luma file.
- One 2,809-row ThinkingAI export is a superseded snapshot and is excluded.
- The analysis uses the remaining 49 event exports: 47 contain data and two are empty.
- 28,195 source rows
- 21,507 invited-only rows excluded
- 6,688 non-invited registrations
- 4,633 unique non-invited email addresses
- 39 display titles and 49 event instances; repeated titles can be distinct events.

One pending application in the network layer was created on October 1. It contains no core research answer. Removing it would reduce both the registration count and the unique-email count by one, but would not change a published question base or finding. All reported research answers were submitted by September 30, 2026.

## What “non-invited” means

“Non-invited” is a source-system category, not an attendance label. It includes approved, declined, pending and waitlisted applications. The 6,688 rows comprise 3,118 approved, 3,085 declined, 414 pending and 71 waitlisted applications. The reports therefore describe applicants and research respondents, not attendees. The `checked_in_at` field is too inconsistently populated to support attendance claims.

## Sample flow by instrument

| Instrument | Non-invited rows | Unique emails | Published question base | Completion from unique emails |
|---|---:|---:|---:|---:|
| Finance, four events | 632 | 547 | 389 | 71% |
| CISO, three events | 234 | 195 | 151 | 77% |
| Marketing status | 252 | 252 | 230 | 91% |
| Marketing impact | 271 | 242 | 218 | 90% |
| YC founder pricing and buyer | 190 | 190 | 148 | 78% |
| Investor ownership | 300 | 300 | 245 | 82% |
| Investor impact | 300 | 300 | 237 | 79% |
| AI Leaders | 314 | 314 | 71–101 | 23–32% |

The AI Leaders instrument has the largest item-level nonresponse. Its 919-row export contains 314 non-invited applications, while individual research questions have 71 to 101 answers. Each statistic therefore carries its own base; 314 is not used as a universal sample size.

## Qualitative open-text review

Open-text responses are used only as event-specific context. They were read for recurring terms and topics without a formal codebook, independent double-coding or inter-rater testing. No theme counts, theme shares or comparisons between events are published.

| Event or instrument | Nonblank responses reviewed | How the reading is used |
|---|---:|---|
| Marketing Measurement and Attribution in the AI Era | 74 | Qualitative context for attribution and measurement questions |
| Partner Marketing and the New AI Ecosystem | 29 | Qualitative context for partner and ecosystem questions |
| Agentic AI Meets Go-to-Market | 123 | Qualitative context for agents, go-to-market work and peer connections |
| AI Leaders | 43 | Qualitative context for adoption, agents, infrastructure, deployment and governance |

The bases above are counts of nonblank responses after the report's source filters. They are not denominators for estimating the prevalence of a theme, and the four sets are not combined into a single marketing or executive distribution.

## Calculation rules

Respondents are deduplicated by normalized email for each question, with the latest answer retained. Multi-select answers use the any-mention convention. A respondent can therefore appear in more than one answer category, and shares can sum above 100 percent. No headline is based on fewer than 40 responses. Bases from 10 to 39 are labeled directional.

The investor-ownership choice “Individual teams, bottom-up” contains a literal comma inside one source-system option. It is preserved as one category before multi-select answers are separated. Its count is 45 of 245 overall; it is not two 45-response categories.

Some multi-select combinations are internally ambiguous because the form allowed an exclusive-sounding choice alongside another answer. These include seven signoff answers naming “no single owner” plus an owner, eleven budget answers naming “no clear budget” plus a funding source, and five investor-impact answers naming both a positive impact and “nothing measurable.” The any-mention method is applied consistently and disclosed; these combinations are not silently recoded.

The title-seat and investor-seniority tables classify self-reported titles with keyword rules. They are analytical cuts, not respondent-selected roles. Categories with fewer than ten respondents are omitted from displayed cuts. The sign-off comparison also focuses on four selected operating groups rather than the large other/unclassified group. Each affected table states how many respondents its shown rows cover.

Title-seat classification uses the following precedence when a title matches more than one category:

1. Finance: CFO, chief financial, senior finance leadership, controller, treasurer or FP&A.
2. Security: CISO, chief information security, senior security leadership or cybersecurity.
3. Marketing: CMO, chief marketing or growth, or senior marketing leadership.
4. Technology: CTO, CIO, chief technology, information, data or AI roles; senior engineering, AI or data leadership; and technical practitioner titles.
5. CEO or founder: CEO, chief executive, founder, owner or president.
6. Investor or partner: partner, investor, venture, capital, investment, principal, managing director or portfolio.
7. Legal: general counsel, chief legal, attorney, lawyer, counsel or legal officer.
8. People: CHRO, chief people, human resources, people officer or talent.
9. Other or unclassified: all remaining supplied titles.

Investor seniority is classified separately. Partner, managing director, founder and C-suite titles form the senior group; vice president, principal, director and head-of titles form the middle group; other nonblank titles form the remaining group. These rules are deterministic but imperfect, so small rows are labeled directional.

The maturity-by-cost and fleet-by-cost tables are matched at the respondent level using normalized email. Their bases differ from the 36-point production-versus-cost-visibility comparison, which uses the full base for each of two separate questions. That difference is descriptive, not a matched-respondent result.

Two additional matched cross-tabs support the full-length founder and investor reports. The founder pricing-by-buying-doorway table contains the 148 respondents who answered both founder questions. The investor ownership-by-portfolio-impact table contains 234 respondents who answered both investor questions. Both source questions are multi-select: each displayed column uses any mention, a respondent can appear in more than one column, and column percentages do not describe mutually exclusive groups. These tables show cross-sectional association only; they do not establish that a buyer, owner or pricing model caused an outcome.

Founder and investor headline lines are carried forward where no newly fielded version of the instrument exists. They are not described as October movement.

Two calculated figures require particular care. The 86 percent positive-impact share counts respondents selecting at least one of three positive portfolio impacts; it does not measure investment conviction. The 34-point security comparison subtracts the share naming a dedicated AI security budget line from the share naming agent access as a problem; it does not measure budget adequacy because respondents could report other funding sources.

Results differ by application status. For example, the CEO appears in 31 percent of approved common-signoff answers and 54 percent of declined answers, while finance appears in 47 percent and 26 percent, respectively. Application status is itself a selection outcome and does not establish attendance, so this comparison is retained as a sensitivity check rather than a headline finding.

## Exhibit library

The original source library contains 25 October-specific PNG files. The corrected publication pack adds `26-board-oversight-signals.png` and uses 25 valid chart assets; `20-control-gaps.png` is obsolete, unreferenced and excluded from the upload pack. Chart 26 is a board-specific synthesis of six already validated survey signals, keeps each numerator or base visible and explicitly does not pool the measures into a score. Each quantitative chart states its question base and source convention. Source-report tables provide the exact distributions behind the individual measures.

## Privacy

The reports and aggregate data contain no identifying respondent information. Free-text revenue, funding and ARR answers are not published.